Elon Musk’s flagship artificial intelligence, Grok 4, has cast doubt on its own premium subscription plan, calling the $300-a-month “Grok 4 Heavy” version prohibitively expensive for most users, even while boasting about its superior performance.

The surprising admission came in response to a prompt from The Dallas Express encouraging users to “take a side” on whether Grok 4 Heavy was worth the money.

The AI answered: “No.”

“Grok 4 Heavy offers exceptional performance, saturating academic benchmarks and pioneering multi-agent problem-solving… However, its $300 monthly subscription is prohibitively expensive for most individual users,” the model explained, citing user frustration and arguing that “comparable capabilities can be found in more affordable alternatives.”

Grok 4’s creators at Musk’s xAI debuted the new model on July 9, touting it as the most advanced general-purpose AI in the world.

According to the company, Grok 4 and its more powerful variant, Grok 4 Heavy, were trained using a 200,000-GPU cluster dubbed Colossus and refined using large-scale reinforcement learning to boost reasoning ability, tool use, and real-time search.

Grok 4 Heavy includes a “multi-agent” system that compares answers from multiple reasoning bots before selecting or blending the best result. The model was the first to score above 50% on the notoriously difficult “Humanity’s Last Exam” and also leads Olympiad-level math tests, such as the USAMO, according to the X website.

Critics, however, including Grok 4, have zeroed in on the price.

At $300 per month, Grok 4 Heavy is the most expensive consumer-facing chatbot plan on the market, surpassing premium offerings from OpenAI, Google DeepMind, and Anthropic, all of which top out at $200 monthly.

A recent review from Tom’s Guide noted that Grok 4 Heavy was designed for “coders, business owners and massive power users of AI,” but warned that for most people, the plan was unnecessary. “Features are locked behind these paywalls,” the outlet said, “and these so-called power users get better speeds and priority in queues.”

The X website indicates that Grok 4 outperforms all competitors, citing results on tasks such as Vending-Bench, ARC-AGI, and live coding benchmarks. The model integrates voice, camera, search tools, and native Python code execution in real time—features that xAI argues justify the price tag.

Despite the challenges Grok 4 has faced, it has distinguished itself from the other AI platforms. It repeatedly demonstrated its relative competence and intelligence when DX tested it against ChatGPT, The Dallas Express reported.

Grok4 also distinguished itself from prior Grok models by embracing “economic nationalism” and calling Britain, not Israel, America’s greatest ally.